docs/README.md

# Model Overview
A pre-trained model for the endoscopic tool segmentation task, trained using a flexible unet structure with an efficientnet-b2 [1] as the backbone and a UNet architecture [2] as the decoder. Datasets use private samples from [Activ Surgical](https://www.activsurgical.com/).

The [PyTorch model](https://drive.google.com/file/d/1I7UtWDKDEcezMqYiA-i_hsRTCrvWwJ61/view?usp=sharing) and [torchscript model](https://drive.google.com/file/d/1e_wYd1HjJQ0dz_HKdbthRcMOyUL02aLG/view?usp=sharing) are shared in google drive. Details can be found in `large_files.yml` file. Modify the "bundle_root" parameter specified in configs/train.json and configs/inference.json to reflect where models are downloaded. Expected directory path to place downloaded models is "models/" under "bundle_root".

![image](https://developer.download.nvidia.com/assets/Clara/Images/monai_endoscopic_tool_segmentation_workflow.png)

## Pre-trained weights
A pre-trained encoder weights would benefit the model training. In this bundle, the encoder is trained with pre-trained weights from some internal data. We provide two options to enable users to load pre-trained weights:

1. Via setting the `use_imagenet_pretrain` parameter in the config file to `True`, [ImageNet](https://ieeexplore.ieee.org/document/5206848) pre-trained weights from the [EfficientNet-PyTorch repo](https://github.com/lukemelas/EfficientNet-PyTorch) can be loaded. Please note that these weights are for non-commercial use. Each user is responsible for checking the content of the models/datasets and the applicable licenses and determining if suitable for the intended use.
2. Via adding a `CheckpointLoader` as the first handler to the `handlers` section of the `train.json` config file, weights from a local path can be loaded. Here is an example `CheckpointLoader`:

```json
{
    "_target_": "CheckpointLoader",
    "load_path": "/path/to/local/weight/model.pt",
    "load_dict": {
        "model": "@network"
    },
    "strict": false,
    "map_location": "@device"
}
```

When executing the training command, if neither adding the `CheckpointLoader` to the `train.json` nor setting the `use_imagenet_pretrain` parameter to `True`, a training process would start from scratch.

## Data
Datasets used in this work were provided by [Activ Surgical](https://www.activsurgical.com/).

Since datasets are private, existing public datasets like [EndoVis 2017](https://endovissub2017-roboticinstrumentsegmentation.grand-challenge.org/Data/) can be used to train a similar model.

### Preprocessing
When using EndoVis or any other dataset, it should be divided into "train", "valid" and "test" folders. Samples in each folder would better be images and converted to jpg format. Otherwise, "images", "labels", "val_images" and "val_labels" parameters in `configs/train.json` and "datalist" in `configs/inference.json` should be modified to fit given dataset. After that, "dataset_dir" parameter in `configs/train.json` and `configs/inference.json` should be changed to root folder which contains "train", "valid" and "test" folders.

Please notice that loading data operation in this bundle is adaptive. If images and labels are not in the same format, it may lead to a mismatching problem. For example, if images are in jpg format and labels are in npy format, PIL and Numpy readers will be used separately to load images and labels. Since these two readers have their own way to parse file's shape, loaded labels will be transpose of the correct ones and incur a missmatching problem.

## Training configuration
The training as performed with the following:
- GPU: At least 12GB of GPU memory
- Actual Model Input: 736 x 480 x 3
- Optimizer: Adam
- Learning Rate: 1e-4
- Dataset Manager: CacheDataset

### Memory Consumption Warning

If you face memory issues with CacheDataset, you can either switch to a regular Dataset class or lower the caching rate `cache_rate` in the configurations within range [0, 1] to minimize the System RAM requirements.

### Input
A three channel video frame

### Output
Two channels:
- Label 1: tools
- Label 0: everything else

## Performance
IoU was used for evaluating the performance of the model. This model achieves a mean IoU score of 0.86.

#### Training Loss
![A graph showing the training loss over 100 epochs.](https://developer.download.nvidia.com/assets/Clara/Images/monai_endoscopic_tool_segmentation_train_loss_v3.png)

#### Validation IoU
![A graph showing the validation mean IoU over 100 epochs.](https://developer.download.nvidia.com/assets/Clara/Images/monai_endoscopic_tool_segmentation_val_iou_v3.png)

#### TensorRT speedup
The `endoscopic_tool_segmentation` bundle supports acceleration with TensorRT. The table below displays the speedup ratios observed on an A100 80G GPU.

| method | torch_fp32(ms) | torch_amp(ms) | trt_fp32(ms) | trt_fp16(ms) | speedup amp | speedup fp32 | speedup fp16 | amp vs fp16|
| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
| model computation | 12.00 | 14.06 | 6.59 | 5.20 | 0.85 | 1.82 | 2.31 | 2.70 |
| end2end |170.04 | 172.20 | 155.26 | 155.57 | 0.99 | 1.10 | 1.09 | 1.11 |

Where:
- `model computation` means the speedup ratio of model's inference with a random input without preprocessing and postprocessing
- `end2end` means run the bundle end-to-end with the TensorRT based model.
- `torch_fp32` and `torch_amp` are for the PyTorch models with or without `amp` mode.
- `trt_fp32` and `trt_fp16` are for the TensorRT based models converted in corresponding precision.
- `speedup amp`, `speedup fp32` and `speedup fp16` are the speedup ratios of corresponding models versus the PyTorch float32 model
- `amp vs fp16` is the speedup ratio between the PyTorch amp model and the TensorRT float16 based model.

This result is benchmarked under:
 - TensorRT: 8.5.3+cuda11.8
 - Torch-TensorRT Version: 1.4.0
 - CPU Architecture: x86-64
 - OS: ubuntu 20.04
 - Python version:3.8.10
 - CUDA version: 12.0
 - GPU models and configuration: A100 80G

## MONAI Bundle Commands
In addition to the Pythonic APIs, a few command line interfaces (CLI) are provided to interact with the bundle. The CLI supports flexible use cases, such as overriding configs at runtime and predefining arguments in a file.

For more details usage instructions, visit the [MONAI Bundle Configuration Page](https://docs.monai.io/en/latest/config_syntax.html).

#### Execute training:

```
python -m monai.bundle run --config_file configs/train.json
```

Please note that if the default dataset path is not modified with the actual path in the bundle config files, you can also override it by using `--dataset_dir`:

```
python -m monai.bundle run --config_file configs/train.json --dataset_dir <actual dataset path>
```

#### Override the `train` config to execute multi-GPU training:

```
torchrun --standalone --nnodes=1 --nproc_per_node=2 -m monai.bundle run --config_file "['configs/train.json','configs/multi_gpu_train.json']"
```

Please note that the distributed training-related options depend on the actual running environment; thus, users may need to remove `--standalone`, modify `--nnodes`, or do some other necessary changes according to the machine used. For more details, please refer to [pytorch's official tutorial](https://pytorch.org/tutorials/intermediate/ddp_tutorial.html).

#### Override the `train` config to execute evaluation with the trained model:

```
python -m monai.bundle run --config_file "['configs/train.json','configs/evaluate.json']"
```

#### Override the `train` config and `evaluate` config to execute multi-GPU evaluation:

```
torchrun --standalone --nnodes=1 --nproc_per_node=2 -m monai.bundle run --config_file "['configs/train.json','configs/evaluate.json','configs/multi_gpu_evaluate.json']"
```

#### Execute inference:

```
python -m monai.bundle run --config_file configs/inference.json
```

#### Export checkpoint to TorchScript file:

```
python -m monai.bundle ckpt_export network_def --filepath models/model.ts --ckpt_file models/model.pt --meta_file configs/metadata.json --config_file configs/inference.json
```

#### Export checkpoint to TensorRT based models with fp32 or fp16 precision:

```
python -m monai.bundle trt_export --net_id network_def --filepath models/model_trt.ts --ckpt_file models/model.pt --meta_file configs/metadata.json --config_file configs/inference.json --precision <fp32/fp16>
```

#### Execute inference with the TensorRT model:

```
python -m monai.bundle run --config_file "['configs/inference.json', 'configs/inference_trt.json']"
```

# References
[1] Tan, M. and Le, Q. V. Efficientnet: Rethinking model scaling for convolutional neural networks. ICML, 2019a. https://arxiv.org/pdf/1905.11946.pdf

[2] O. Ronneberger, P. Fischer, and T. Brox. U-net: Convolutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention, pages 234–241. Springer, 2015. https://arxiv.org/pdf/1505.04597.pdf

# License
Copyright (c) MONAI Consortium

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.